Regulation
AI giants promise Trump they will police themselves, with no law to make them
Executives from Google, Anthropic, Meta, OpenAI and Nvidia signed a voluntary safety agreement at the White House. Trump calls it “morally binding”, and it is not clear who will check that it is kept.
The essentials
3 confirmed facts · 2 according to the source · 3 open questions
- Executives from Google, Anthropic, Meta, OpenAI and Nvidia, among other companies, signed an agreement on AI safety at the White House on Tuesday.234
- The text calls for internal controls, an in-house team to review them, an outside evaluator and a board committee to receive the reports.34
- The agreement is voluntary and imposes no legal obligations. Trump described it as “morally binding”.345
- According to the sourceAccording to WIRED, a spokesperson for the FTC, the US consumer protection agency, confirmed that there is an investigation into AI companies, without giving details.3
- According to the sourceAccording to the Wall Street Journal, cited by Xataka, several executives criticised Dario Amodei, of Anthropic, for being alarmist in public about the risks.4
Why it matters
The safety of the most advanced models is left in the hands of those who build them. The agreement creates no legal obligations and, according to the New York Times, cited by Xataka, it is not clear who would monitor compliance4. WIRED notes that it is not the broad regulation that people inside the companies themselves are asking for3.
The only route to enforcement in sight is the FTC, and it looks narrow: according to WIRED, it can act if a company breaks its promises, not decide what is safe3. If the investigation moves forward, it could offer clues about how far that route goes, although the FTC has not said what it is investigating3. Our reading is that, for now, there are public promises and no rules.
The details
The document is called “The White House Accord on Super Intelligence” and was announced on Tuesday after a lunch with Donald Trump, according to a WIRED report of 30 September3. It orders nothing: it says companies should apply the measures3. Ars Technica speaks of two dozen signatory companies, which will also meet regularly to set common safety standards2. Xataka adds that it runs to two pages and sees a case for turning it into law only “over time”4.
And if a company fails to comply? According to WIRED, breaking a public promise can be considered a deceptive practice, and that is where the FTC, the US agency that protects consumers, could act; that is the view of Neil Chilson, a former chief technologist at the agency3. The same outlet notes that the usual remedy in such cases is to make the company promise not to lie again3.
After the lunch, the New York Post reported that the FTC is preparing a broad investigation into Anthropic, OpenAI, other labs and METR, a non-profit organisation that evaluates models from the outside3. According to WIRED, an agency spokesperson confirmed that an investigation exists, without specifying which consumer issues concern it3. According to the Post, it has not yet sent formal requests for information3.
The show of unity had cracks. According to the Wall Street Journal, cited by Xataka, several executives, among them Jensen Huang, of Nvidia, asked Dario Amodei why he is so alarmist in public; he replied that the industry owes people honesty4. Nor do accounts agree on who attended for OpenAI: Ars Technica lists Sam Altman among the signatories2, and the WIRED podcast places Greg Brockman, its president, at the lunch, and not Altman5.
What we don't know
- The full list of signatories and who represented OpenAI: Ars Technica names Sam Altman as a signatory, and the WIRED podcast places Greg Brockman at the lunch.
- What exactly the FTC is investigating and which companies: the agency has given no details.
- Who will check that the companies keep to what they signed and what the consequences of not doing so would be.
Background
These companies had already signed similar voluntary commitments with other governments. According to WIRED, in early 2025 the United Kingdom and the Republic of Korea announced a set of “Frontier AI Safety Commitments” with promises of internal and external testing of models3. The New York Times, cited by Xataka, notes that the voluntary commitments of the Biden era have largely been forgotten4. Ars Technica places the agreement in the middle of a run of safety incidents, which have led OpenAI to stop training runs and pause launches2.
Sources
- 1Here’s what AI leaders are saying about Trump’s new safety plan
- 2Trump plan to combat AI risks hinges on Big Tech pals policing themselves
- 3Trump’s AI Safety ‘Accord’ Is a Fancy Pinky-Swear
- 4Jensen Huang y otros CEO han confrontado a Dario Amodei por sus mensajes tan extremos en materia de seguridad
- 5Trump’s ‘Morally Binding’ AI ‘Accord,’ the Rise of AI Agents, and Extremists on the Ballot
- 6Whatever AI Safety Is, It’s Not This
Was this story useful? Yes Not really
This story was written in Spanish by an artificial intelligence system from the sources listed above, and translated automatically. Before publication, a program checks that every figure and every name appears in the sources, and that the translation keeps the same facts, figures and sources. It has not been reviewed by a person. Spanish original · How we work (in Spanish)